Artificial selection for determinate growth habit in soybean

Identification of Soybean Genes Homologous to Arabidopsis TFL1.

The Arabidopsis TERMINAL FLOWER1 (TFL1) gene was previously identified by isolation of the recessive mutations tfl1 in the TFL1 gene by screening a M2 population derived from EMS-mutagenized seeds of ecotype Columbia (22). The recessive mutations resulted in the conversion of the normally indeterminate inflorescence to a determinate inflorescence condition (22⇓–24). By BLAST searching Arabidopsis TFL1 against the soybean (c.v., Williams 82) whole genome sequence (25), we identified four soybean gene models, Glyma03g35250.1, Glyma10g08340.1, Glyma13g22030.1, and Glyma19g37890.1, that are homologous to TFL1 (Fig. S1). Phylogenetic analysis of these genes suggest that Glyma03g35250.1/Glyma19g37890.1 and Glyma10g08340.1/Glyma13g22030.1 are two homeologous pairs, pre-sumably derived from the soybean genome duplication event that occurred ~50 million years ago (MYA) (25, 26), whereas the two members of each pair likely resulted from the more recent soybean genome duplication event (i.e., allotetraploidization) (27) that took place ~13 MYA (Fig. S1) (26).

The Dt1 locus of soybean was recently fine-mapped as a major quantitative trait locus between two simple sequence repeat (SSR) markers, Sat_099 and Satt229, on LG L (7), which is now designated as chromosome 19 (Gm19) (25). We anchored the SSR markers to the Gm19 sequence and found that Glyma19g37890.1 (designated as GmTfl1) was one of the 380 annotated genes physically located between Sat_099 and Satt229 (24) (Fig. 1). This suggests that GmTfl1 may be a candidate gene for Dt1. Because Williams 82 is a typical indeterminate cultivar, it is likely that GmTfl1 is the candidate Dt1 allele.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Anchoring genetic markers to the genomic sequence to define the candidate Dt1 gene. Vertical bar between Sat_099 and Satt006 on the genetic map and vertical bar on LG and chromosome sequence indicate the candidate Dt1 gene, Glyma19g37890.1. Gene model was predicted and is depicted by the cartoon underneath the “chromosome.”

Allelic Variation of the Candidate Gene in the Wild and Cultivated Soybean Populations.

In an attempt to address whether GmTfl1 and the mutations, if any, that may have occurred in this gene are responsible for the conversion from an indeterminate to determinate phenotype observed in many soybean cultivars, we first sequenced the GmTfl1 locus in a wild G. soja population and three soybean populations, representing genotypic groups that likely existed before and after genetic bottlenecks (e.g., domestication to produce landraces, introduction of relatively few landraces to North America, and selective breeding) (28). Fourteen unique SNPs and two insertions/deletions (indels) were detected (Table S1). Of the 14 SNPs, 10 were found in noncoding regions and four in exons. Interestingly, each of the four exonic SNPs generated a single amino acid nonsysnonymous substitution (Table S1). Not a single individual genotype was found to contain more than one unique amino acid substitutions. Compared with the Williams 82 reference GmTfl1 sequence, these four amino acid substitutions (referred as to Gmtfl1) were only detected in the cultivated soybeans, whereas G. soja genotypes were identical to Williams 82.

Association Between Determinacy and Allelic Nonsynonymous Mutations.

To elucidate whether GmTfl1 is Dt1, and whether any or all of the mutations caused the transition from indeterminate type to determinate phenotype, we conducted an association analysis between the mutations and phenotypes using the three aforementioned soybean populations. The stem growth habit phenotypes of the soybean cultivars in these populations were obtained from the USDA Soybean Germplasm Collection database at National Plant Germplasm System (NPGS) (http://www.ars-grin.gov/npgs/), and some of them were directly examined in this study. Of the 89 soybean cultivars, 39 are indeterminate, 41 are determinate, and nine are semideterminate. We found that each of the 39 indeterminate cultivars exhibited the same amino acid sequence as encoded by GmTfl1 in Williams 82, whereas none of the 41 determinate cultivars contain the Williams 82 amino acid sequence but instead possess one (or another) of the four amino acid substitutions (Fig. 2A and Table S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Inferring the candidate gene by association analysis. (A) Distribution and association of four independent Gmtfl1 mutations with determinacy in four wild and cultivated soybean populations. The genetic structure of the populations was depicted by the vertical bars along the horizontal axis, in which the proportions of ancestry that can be attributed to each cluster were indicated by the length of each colored segment. The GmTfl1/GmTfl1 or Gmtfl1/Gmtfl1 genotypes of individual cultivars were marked by thin vertical bars above the plot of population structure, and their phenotypes, i.e., indeterminacy (I), semideterminacy (S), and determinacy (D), were indicated by up triangles, diamonds and down triangles, respectively. (B) Alignment of the amino acid sequences encoded by the GmTfl1 and Gmtfl1 alleles showing four single amino acid substitutions caused by four corresponding point mutations.

Two semideterminate cultivars in this study were found to have the GmTfl1 allele. A previous study demonstrated that the genotype of Dt1/Dt1 ordinarily displays an indeterminate phenotype, but in the presence of Dt2, a dominant allele at another locus controlling stem growth habit, the Dt1/Dt1;Dt2/Dt2 genotype will display semideterminate phenotype (6); thus, these two semideterminate cultivars were assumed to contain both Dt1 and Dt2 alleles (Fig. 2A and Table S2). The other seven semideterminate cultivars were found to have Gmtfl1 allele. Because it is generally difficult to precisely define the semideterminate phenotypes (6), these nine cultivars were not included in the association analysis below.

G. soja accessions are typically viny, and highly diverged in plant architecture and morphology from G. max; thus, the stem growth habit of the G. soja accessions included in this study was not carefully measured. All of the 20 G. soja accessions were found to contain the same GmTfl1 genotype as Williams 82, which seemingly associates the indeterminacy with G. soja as is generally conjectured.

Thus, we observed a perfect association between the amino acid substitutions and the determinacy when G. soja accessions and the semideterminate cultivars were excluded. This suggests that GmTfl1 is the Dt1 allele, and the four single-point mutations (which could be characterized as functional SNPs) resulted in the four distinct amino acid substitutions are dt1 alleles.

Excluding the four functional SNP variants, the GmTfl1 alleles in the four populations were classified into two distinct types, designated as GmTfl1-a and GmTfl1-b. The four mutations were subclassifiable as Gmtfl1-ta, Gmtfl1-bb, Gmtfl1-tb, and Gmtfl1-ab (Fig. 2B). GmTfl1-a and Gmtfl1-ta share the same form, and GmTfl1-b, Gmtfl1-bb, Gmtfl1-tb, and Gmtfl1-ab share the other form, suggesting that Gmtfl1-ta was derived from GmTfl1-a whereas Gmtfl1-bb, Gmtfl1-tb, and Gmtfl1-ab were derived from GmTfl1-b. Linkage disequilibrium (LD) analysis showed that the SNPs and indels in the first intron (from +285 to +311) are linked with the two SNPs in the 5′UTR (–499 and –410), but the four functional SNPs did not show LD with the other sites (Fig. S2A). We also sequenced Glyma19g37900.1, a gene flanking GmTfl1, in six landraces that contain different alleles at the GmTfl1 locus, and found that LD exists between the SNPs detected at the Glyma19g37900.1 locus and the nonfunctional polymorphisms (–499, –410, and +285 to +311) at the GmTfl1 locus (Fig. S2C). These observations indicate that the transition from indeterminate type to determinate type was not caused by the linked polymorphisms within the GmTfl1 locus, or between the GmTfl1 locus and its flanking gene, but by the four functional mutations. These observations further strengthen the inference that that GmTfl1 is Dt1.

In addition to the four populations analyzed above, we sequenced the GmTfl1 locus in 17 previously described determinate soybean cultivars (Table S3). All of the 17 cultivars were found to be Gmtfl1 mutations (two Gmtfl1-ta, two Gmtfl1-tb, and 13 Gmtfl1-ab), a result consistent with the association analysis above. We also sequenced the GmTfl1 locus in three semideterminate isogenic lines that share the Clark (an indeterminate cultivar) genetic background but differ from Clark at the Dt2 locus (6), and did detect the GmTfl1 allele in all these isogenic lines (Table S3).

GmTfl1 Complements the Functions of TFL1 in Arabidopsis.

To validate the function of GmTfl1 for indeterminacy (vs. Gmtfl1 for determinacy) we introduced the Williams 82 GmTfl1 allele into the Arabidopsis determinate mutant (tfl1-1) (24) (Materials and Methods), and obtained two transgenic lines, one of which is shown in Fig. 3C. The absence of Arabidopsis TFL1 allele and the presence of the soybean GmTfl1 allele in the transgenic lines were confirmed by PCR analysis and sequencing of PCR fragments (Fig. 3 G and ⇓I). The transgenic (GmTfl1) lines (Fig. 3 C and G) showed the same phenotypes as the wild-type Arabidopsis (i.e., indeterminate and late flowering). Because the transgene (GmTfl1) is a combination of the Arabidopsis TFL1 promoter and the protein coding sequence (CDS) of the GmTfl1 allele, the conversion of the transgenic lines from the mutant type (determinate and early flowering) to the wild type would be interpreted that the transgene in the Arabidopsis (tfl1/tfl1) mutant fully complements the functions of TFL1 observed in the wild-type Arabidopsis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Functional analysis of GmTfl1 and Gmtfl1 alleles in the Arabidopsis tfl1 mutants. (A) Wild-type Arabidopsis (TFL1/TFL1), (B) tfl1-1 mutant, (C) tfl1-1 mutant with transgene GmTfl1. (D) tfl1-1 mutant with transgene Gmtfl1-ab. (E–H) Cartoons of growths of the wild-type (TFL1), tfl1 mutant, the GmTfl1 transgenic line, and Gmtfl1-ab transgenic line, as shown in A, B, C, and D, respectively. Curves and letters beneath the cartoons illustrate a single nucleotide difference (G and A) between Arabidopsis TFL1 and tfl1-1 alleles detected in the three lines by sequencing. (I and J) Confirmation of presence of soybean GmTfl1 and Gmtfl1-ab alleles, marked by a single nucleotide (A and T), respectively, in the transgenic Arabidopsis tfl1 lines by PCR and sequencing of PCR fragments.

The question remained whether the nonsynonymous substitutions (Gmtfl1 alleles) detected at the GmTfl1 locus in the cultivated soybean have no or diminished functions relative to the GmTfl1 allele for indeterminacy. To address this question, we introduced the Gmtfl1-ab, the predominant allele detected in the cultivated soybean populations (Fig. 2A), into the Arabidopsis tfl1-1 mutants (Materials and Methods), and obtained eight transgenic (Gmtfl1) lines. The absence of the Arabidopsis TFL1 allele and the presence of the soybean Gmtfl1-ab allele in the eight transgenic lines were confirmed by PCR analysis and sequencing of PCR fragments (Fig. 3 H and J). We found that each of the eight lines showed phenotypes nearly identical to that of the Arabidopsis tfl1-1 mutant. The phenotypes of one of the eight lines are illustrated in Fig. 3 D and H.

Evolutionary Diversification between GmTfl1 and its Homeolog.

Since GmTfl1 and Glyma03g35250.1 are thought to be a homeologous pair (Fig. 1B), it would be interesting to track the evolutionary difference between GmTfl1 and Glyma03g35250.1. We thus sequenced the Glyma03g35250.1 locus in the same populations used for the analysis of the GmTfl1 locus. Five SNPs were detected at the Glyma03g35250.1 locus in the G. soja population, but none were found in the cultivated populations (Table S1 and Fig. S2B). The level of nucleotide diversity at both GmTfl1 and Glyma03g35250.1 loci (Table 1) is lower than the average in the G. soja population estimated based on a set of gene fragments (28). In addition, nonsynonymous substitutions were not found at either locus in the G. soja population (Table 1), suggesting that both genes have undergone purifying selection. However, GmTfl1 and Glyma03g35250.1 exhibited a substantial difference in diversity in the cul-tivated soybean populations. For example, the Glyma03g35250.1 allele was invariant among all of the members of the soybean landrace population, whereas the nonsynonymous substitutions at the GmTfl1 locus that resulted in the four Gmtfl1 alleles were observed in the same population (Table 1). Together, these observations suggest that the fixation of the four Gmtfl1 alleles in cultivated determinate soybean would be the outcome of deliberate human selection during the development of soybean landraces.

Differential Expression of GmTfl1 and its Homeolog.

To shed lights on the functional diversity of GmTfl1 and Glyma03g35250.1, we compared their expression pattern. Quantitative RT-PCR was used to profile the expressions of GmTfl1 and Glyma03g35250.1, in the indeterminate cultivar Williams 82 in different tissues and at different developmental stages, GmTfl1 was mainly expressed in young roots, young leaves and flowers seven day after flowering (7DAF), whereas Glyma03g35250.1 was mainly expressed in young roots, young stems and buds (Fig. S3). Given that Arabidopsis TFL1 is involved in inflorescence meristem development pathway (22, 24), high-level of expression of GmTfl1 in flowers 7DAF is expected. Thus, the lack of expression of Glyma03g35250.1 at this stage may be considered as evidence that the Glyma03g35250.1 was subfunctionalized or neo-functionalized. This inference is echoed by the analysis of allelic variation at both GmTfl1 and Glyma03g35250.1 loci in the soybean populations. It can be deduced that neither Glyma03g35250.1 nor the other pair of homeologous genes (Glyma13g22030.1, Glyma10g08340.1), homologous to TFL1, are potential candidates for the Dt2 locus, as these three loci are not located on LG G (chromosome 18), where the Dt2 was mapped.

Timing and Nature of Artificial Selection for the Gmtfl1 Alleles.

None of the four Gmtfl1 alleles identified in G. max were detected in the G. soja population analyzed in this study. To search for evidence of the history of the human selection with respect to the Gmtfl1 alleles, we sequenced the GmTfl1 locus in a minicore collection of 195 soybean landraces, which were selected based on the genetic structure of a core collection of 1,863 landraces that maximally represent the 23,587 Chinese soybean landraces deposited in the Crop GenBank at the Chinese Academy of Agricultural Sciences. We subsequently analyzed the distribution of the four Gmtfl1 alleles in the core collection of landraces with respect to their genetic diversity and geographic distribution. The Gmtfl1 alleles were seen in all of the major branches of the Neighbor-Joining tree of the 195 landraces constructed based on 59 SSR markers (3) (Fig. 4A). It is noticeable that Gmtfl1-ta and Gmtfl1-tb were found in a highly diverged group of (seven) landraces that are the most closely related to G. soja, a wild accession used as an outgroup, and six of these seven landraces show “semi-wild” phenotypes, such as viny stems and dark brown seed coat (3). These data indicate that the human selection for determinacy must have occurred before the radiation of all of the lineages of these Chinese landraces, either just after or during the major domestication transition. Although the Gmtfl1 landrace alleles were found in all of the three large soybean-growing ecological regions, referred to as Northern eco-region (NR), Huang-Huai eco-region (HR), and Southern eco-region (SR), which were subclassified into NESp and NSp, HSp and HSu, and SR, CSp, SSp, SAu, and SSu subregions, respectively (3), the four Gmtfl1 alleles are mainly present in the landraces distributed in SR, GmTfl1 is mainly found in the NR, and GmTfl1 and Gmtfl1 are nearly equally distributed in HR (Fig. 4B, Table S4, and Fig. S4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Allelic mutations at the GmTfl1 locus in the context of genetic diversity and eco-geographic distribution of a core collection of soybean landraces. (A) Phylogenetic relationship of the landraces assessed by 59 SSR markers and the types of alleles (GmTfl1 or Gmtfl1) detected in individual landraces. (B) Geographical distribution of the landraces in the soybean growing eco-regions or subregions in China.